银行级安全金库 Bank-Grade Security Vault
私钥，从来就不存在。 There is no private key. Ever.
Alph AI · MPC Keyless 安全体系 Alph AI · MPC Keyless Security System
If it breaks, we pay.
If it breaks, we pay.
Alph AI 将钱包安全提升为系统级能力：MPC 无私钥、零攻击面隔离、 AWS Nitro Enclaves 与全链路风控，形成可验证、可审计、可复用的银行级安全标准。
黑客无法偷走不存在的东西。
Alph AI elevates wallet security to a system-grade capability: MPC keyless, zero-attack-surface isolation, AWS Nitro Enclaves, and full-path risk control—bank-grade, verifiable, auditable, reusable.
Hackers can’t steal what never exists.
安全架构总结 Security Summary
“There is no private key. Ever.”
私钥从未完整出现，攻击面从根源消失。资产安全不依赖单点信任，而依赖可验证的密码学与物理隔离。
Private keys never fully exist; the attack surface disappears at the root. Asset security relies on verifiable cryptography and physical isolation, not single-point trust.
七层防御 7-Layer Defense
Alph.ai 的七层纵深防御 Alph.ai 7-Layer Defense
MPC 3-of-3 / 2-of-3
“私钥不出现 (Keyless)”：私钥被粉碎成 3 片，签名只需 2 片参与；任何时刻世界上都不存在完整私钥。
“Keyless” design: the key is split into 3 shards, signing needs only 2; the full key never exists.
“黑客无法偷走不存在的东西。”
“Hackers can’t steal what never exists.”
隔离签名服务 Isolated Signing Service
“零外部攻击面 (Zero Attack Surface)”：签名服务运行在完全隔离的计算网络，没有任何外部入口可供黑客入侵。
“Zero Attack Surface”: signing runs in fully isolated networks with no external entry points.
“不仅是防守，是根本没有路。”
“Not just defense—there is no path in.”
物理 / 网络级隔离 Physical / Network Isolation
“独立安全域”：每个节点互不信任、物理隔离。即使攻破一个，也无法影响全局。
Independent security domains with physical isolation. Compromise in one node can’t spread.
“互不信任，才是最大的信任。”
“Mutual distrust is the strongest trust.”
TEE（AWS）
“银行级隔离金库”：敏感计算仅在 AWS Nitro Enclaves 内完成，外部系统无法读取。
“Bank-grade isolated vaults”: sensitive computation happens only inside AWS Nitro Enclaves.
“隔离域内可验证执行。”
“Verifiable execution in isolated enclaves.”
多因子密钥管理 Multi-Factor Key Control
“无法单点突破”：没有单把钥匙可以解密。黑客需同时攻破多个团队、多个系统、多层加密。
No single key can unlock. Attackers must compromise multiple teams, systems, and layers.
“攻击难度呈指数级上升。”
“Attack difficulty grows exponentially.”
签名链路验证 Signature Path Verification
“全路径鉴权”：每一次签名请求都经过完整的防伪造、防重放验证。
End-to-end auth for every signing request with anti-forgery and anti-replay checks.
“每一笔交易都有‘身份证’。”
“Every transaction has an ID.”
Zero-Trust 运维 Zero-Trust Ops
“多人共管 (MPC for Ops)”：没有内部人员（包括创始人）可以单独触碰关键环节。一切皆有审计。
MPC for ops: no single insider can access critical steps. Everything is audited.
“去绝‘内鬼’作恶的可能。”
“Eliminate insider abuse.”
核心理念：让攻击成本从“加法”变为“乘法”。任何单层失守都不足以触达资产本体。
Core idea: turn attack cost from addition to multiplication. No single layer compromise reaches assets.
无私钥金库 Keyless Vault
MPC / TSS 钱包体系 MPC / TSS Wallet System
分布式密钥生成与阈值签名协作，让攻击成本从“加法”变为“乘法”。完整私钥从未出现， 任何单点泄露都无法构成真实风险。
Distributed key generation and threshold signing turn attack cost from addition to multiplication. The full private key never exists; single-point leaks are ineffective.
密钥分片 Key Sharding
私钥碎片化，完整密钥从未出现。
Key shards only; the full key never exists.
门限签名 Threshold Signature
2-of-3 / 3-of-3 门限协作签名。
2-of-3 / 3-of-3 threshold co-signing.
安全域隔离 Security Domains
多域隔离 + 无公网签名节点。
Multi-domain isolation + no public signing nodes.
分布式密钥生成（示意） Distributed Key Generation (Illustration)
传统方案：生成完整私钥 → private_key = generate_private_key()
MPC 方案：分布式生成分片 → shard_1, shard_2, shard_3 = generate_shards_in_distributed_manner()
关键：没有任何设备/人见过完整私钥。
Traditional: generate full private key → private_key = generate_private_key()
MPC: distributed shard generation → shard_1, shard_2, shard_3 = generate_shards_in_distributed_manner()
Key point: no device/person ever sees the full private key.
TEE 安全区 TEE Enclave
可信执行环境（TEE） Trusted Execution Environment (TEE)
分片解密与签名计算仅在 Enclave 内完成。远程证明确保运行环境真实可信，避免 内存提取与供应链攻击风险。
Shard decryption and signing occur only inside the enclave. Remote attestation verifies integrity and mitigates memory extraction and supply-chain risks.
- AWS Nitro Enclaves（隔离执行环境）AWS Nitro Enclaves
- 内存加密 + 密封存储 + 远程证明Memory encryption + sealed storage + remote attestation
- 跨平台安全计算基，降低单点依赖Cross-platform secure base to reduce single-vendor dependency
零信任基础设施 Zero-Trust Infrastructure
零信任基础设施 Zero-Trust Infrastructure
零信任运维原则 Zero-Trust Ops Principles
- 四眼原则：关键操作需至少两人独立验证Four-eyes rule: critical actions require two independent approvals.
- 时间限制访问：临时权限自动过期Time-bound access: temporary privileges auto-expire.
- 行为基线监控：异常操作实时告警Behavioral baselines: anomalies alert in real time.
网络拓扑 Network Topology
网络拓扑：零攻击面设计 Network Topology: Zero Attack Surface
交易前风险扫描 Pre-Trade Risk Scan
合约风险预检流水线 Contract Risk Pre-Check Pipeline
在您输入合约地址的毫秒级窗口内，系统并行完成规则引擎、符号执行与 AI 语义分析， 输出结构化风险报告与操作建议。
Within milliseconds after you input a contract address, the system runs rules, symbolic execution, and AI semantic analysis in parallel to produce a structured risk report and recommendation.
- 地址输入 → 拉取字节码
- 多引擎并行分析（< 3 秒）
- 规则引擎：匹配 300+ 风险模式
- 符号执行：探测隐藏逻辑路径
- AI 语义分析：理解合约真实意图
- 输出结构化报告
- 风险评分（0-100）
- 关键风险项（买卖税、黑名单等）
- 建议操作（禁止/警告/放行）
- Address input → fetch bytecode
- Parallel analysis (< 3s)
- Rules engine: match 300+ risk patterns
- Symbolic execution: detect hidden logic paths
- AI semantics: infer real contract intent
- Structured report
- Risk score (0–100)
- Key risks (taxes, blacklist, etc.)
- Recommendation (block / warn / allow)
执行模拟 Execution Simulation
交易中执行模拟 In-Trade Execution Simulation
交易中阶段对关键场景并行模拟，评估滑点承受与夹击风险，避免执行失败。
Simulate critical scenarios in parallel to assess slippage tolerance and sandwich risk to avoid failed executions.
- 滑点冲击：价格波动容忍度测试Slippage shock: price impact tolerance
- Gas 波动：成本可控性验证Gas volatility: cost control check
- MEV 夹击：抗抢跑能力模拟MEV sandwich: anti-front-running simulation
- 合约回调：重入攻击防护Callback: re-entrancy defense
- 极端波动：断崖式下跌风险Extreme moves: cliff-drop risk
交易后监控 Post-Trade Monitoring
交易后持仓与流动性监控 Post-Trade Holdings & Liquidity Monitoring
交易完成只是风险开始。系统持续监控持仓资产、LP 变化与权限异动， 发现异常即推送预警并提供止损与回放。
Execution is the start of risk. We continuously monitor holdings, LP changes, and permission changes, triggering alerts and providing stop-loss and replay tools.
- LP 异常波动与撤池预警LP volatility & withdrawal alerts
- 顶级持仓地址异动检测Top holder movement detection
- 权限变更与合约升级监控Permission changes & upgrades
- 一键止损 + 风险报告回放One-click stop loss + risk replay
MPC × 风控引擎 MPC × Risk Engine
与 MPC 签名的联动拦截 MPC Signing Interlock
风控系统与 MPC 签名层深度耦合，形成“检测—拦截—审计”的闭环。 高风险交易可直接阻断签名请求，避免资金被“合法转移”。
Risk controls are tightly coupled with MPC signing to form a detect–block–audit loop. High-risk transactions can be denied before signing.
- 风险评分 < 30：拒绝签名Risk score < 30: deny signing
- 确认 Rug 风险：触发临时冻结Rug confirmed: trigger temporary freeze
- 拦截记录进入零信任审计体系Blocks logged into zero-trust audit
审计与透明度 Audit & Transparency
审计与透明度体系 Audit & Transparency
所有签名请求、TEE 远程证明与网络元数据均可审计。结合季度渗透测试与协议形式化验证， 建立可持续的安全透明度机制。
All signing requests, TEE attestations, and network metadata are auditable. Quarterly penetration testing and protocol verification sustain transparency over time.
- 实时技术审计：签名日志 + 远程证明Real-time audit: signing logs + attestations
- 季度第三方安全测试与形式化验证Quarterly third-party tests & formal verification
- 公开透明度报告与漏洞赏金进展Public transparency reports & bug bounty updates
安全团队 Security Team
银行级安全团队与背景 Bank-Grade Security Team
Alph AI 的安全金库由具备传统金融 + Web3 双重背景的安全团队设计与维护。
The security vault is designed and maintained by a team spanning TradFi and Web3 security expertise.
核心负责人 Core Lead
安全团队由 CTO Evan 领导，主导多起关键审计与高危漏洞修复，并在国际安全会议分享链上签名实践。
Led by CTO Evan, the team has led key audits, fixed critical vulnerabilities, and presented on on-chain signing.
- 曾主导 CertiK 多个关键审计项目Led major CertiK audits
- 发现并修复多个高危漏洞Fixed multiple critical issues
- 分享链上签名安全实践Shared on-chain signing practices
跨领域背景 Cross-Disciplinary
团队核心成员覆盖传统金融与 Web3 安全机构，具备 150+ DeFi / Layer1 / 跨链桥审计经验。
Core members span TradFi and Web3 security firms with 150+ DeFi, Layer1, and bridge audits.
- 来自顶级投行/支付机构风控团队From top banks / payments risk teams
- 熟悉 SOC2、ISO 27001 等审计规范SOC2 / ISO 27001 expertise
- 来自 CertiK、Quantstamp、PeckShieldFrom CertiK, Quantstamp, PeckShield
战略合作 Strategic Partner
与拥有 8 年经验的 Bitrue 安全团队建立合作，进一步增强链上资产保护能力。
Partnered with Bitrue’s 8-year security team to strengthen on-chain asset protection.
我们不仅在构建交易平台，更是在打造银行级安全金库。
We’re building a bank-grade security vault, not just a trading platform.
Alph Video Alph Video
私钥，从来就不应该存在。
Private keys should never exist.
MPC 如何让黑客无法攻破
How MPC makes hacks impossible